Products · QNAP · NAS
QNAP QAI-16700
- Processor
- AMD Ryzen 7000 series (desktop, incl. PRO 7645 / 7745)
- Memory (RAM)
- 32 or 192 GB
- Released
- 2026
Configurations
Pick one to see which models it runs. More memory lets larger models fit; bandwidth sets how fast they answer.
| Configuration | Memory | Usable for AI | Bandwidth | LLMs that run well | Launch price |
|---|---|---|---|---|---|
| AMD Ryzen 7000 series (desktop, incl. PRO 7645 / 7745) | 32.0 GB | 26.2 GB | 83 GB/s | 365 of 622 | — |
| AMD Ryzen 7000 series (desktop, incl. PRO 7645 / 7745) · upgraded to the maker's maximum | 192 GB | 173 GB | 83 GB/s | 400 of 622 | — |
Specifications 32.0 GB configuration
| Maker | QNAP |
|---|---|
| Type | NAS |
| Processor | AMD Ryzen 7000 series (desktop, incl. PRO 7645 / 7745) NAS (32 GB) |
| Memory (RAM) | 32.0 GB |
| Usable for AI models | 26.2 GB |
| Memory bandwidth | 83 GB/s |
| FP16 compute | 0.7 TFLOPS |
| Runs models on | CPU cores |
| Drive bays | 12 x 3.5-inch SATA 6Gb/s, 3Gb/s + 4 x 2.5-inch U.2 NVMe Gen4 x2 slots |
| NAS system | QuTS hero |
| Containers (Docker) | Yes |
| Virtual machines | Yes |
| Memory upgrade | UDIMM DDR5 · 4 slots · up to 192 GB per the maker · ECC |
| Graphics card options | PCIe: 3 Slot 1: Gen 4 x 8 or Gen 4 x 4. Maker-supported graphics cards: NVIDIA RTX PRO 6000 Blackwell 96 GB, RTX PRO 5000 Blackwell 48 GB, RTX PRO 4500 Blackwell 32 GB, RTX PRO 4000 Blackwell 24 GB, RTX PRO 2000 Blackwell SFF 16 GB; Intel Arc Pro B70 32 GB listed as…. |
| Notes | AMD Ryzen 9 7900 (12-core/24-thread, up to 5.4 GHz). QNAP itself positions this model for on-premises LLM inference with NVIDIA RTX GPUs. x86-64 Ryzen 9 7000 series CPU, up to 192 GB RAM, Container Station (Docker) and Virtualization Station. |
| Product page | www.qnap.com (opens in a new tab) |
Running AI on the QNAP QAI-16700
Yes, it runs AI models on the NAS itself. Models run on its CPU cores, and this configuration leaves about 26.2 GB for them after the NAS system and its storage services. The grades below assume the NAS isn’t busy with backups or media work at the same time. More memory is the upgrade that matters most: see the configurations.
Apps you can install to run models
| App | How it installs |
|---|---|
| Ollama + Open WebUI | In a container (Docker) |
| Ollama | In a container (Docker) |
| Open WebUI | In a container (Docker) |
| AnythingLLM | In a container (Docker) |
| RAGFlow | In a container (Docker) |
AI features from QNAP
| Feature | What it does | Where the AI runs |
|---|---|---|
| Qsirch RAG Search (cloud LLM mode) | Natural-language search and Q&A over NAS files; retrieval runs on the NAS but the LLM is a cloud API (Gemini, OpenAI, Azure OpenAI or OpenAI-compatible). | Cloud service or another computer |
| Qsirch with RAG (on-premises GPU) | Qsirch full-text search plus RAG linked to a local LLM, with search and generation both on the NAS (maker product page). Needs an NVIDIA GPU in the PCIe slot; the spec-sheet SKU QAI-16700-R9-32G lists no bundled GPU. | On the NAS |
| Container Station AI templates (Ollama, Open WebUI, AnythingLLM, RAGFlow, Dify, Flowise… | One-click local LLM, RAG, image-generation, speech and code-assistant stacks on the built-in GPU. | On the NAS |
| QuMagie + QNAP AI Core | Photo management with on-device face, object and scene recognition (QNAP AI Core image-recognition engine). | On the NAS |
| Qsirch AI Semantic Search | Natural-language/semantic file and image search computed locally; QNAP requires a 64-bit x86 NAS with at least 8 GB RAM. | On the NAS |
| Qsirch on-premises RAG / LLM Core | Runs open models (DeepSeek R1, Gemma 3, gpt-oss 20B, Mistral 7B, Phi-4) locally for RAG search. Requires an NVIDIA GPU assigned to Container Station; QNAP states CPU-only is not supported. Needs QuTS hero h5.2.1+ and Qsirch 6.0.0+. | On the NAS |
Best LLMs for the QNAP QAI-16700 32.0 GB
| Quality | Model | Quant | Memory | Speed | Runs |
|---|---|---|---|---|---|
| 88Excellent quality (A+) · 88 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking. | Qwen3.6 35B-A3B | Q4_K_M | 24.8 GB | 26.7 tok/s | Runs welltight fit |
| 88Excellent quality (A+) · 88 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking. | MiniCPM-V 4.7 35B-A3B | Q4_K_M | 24.3 GB | 26.7 tok/s | Runs welltight fit |
| 88Excellent quality (A+) · 88 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking. | Qwen3.5 35B-A3B | Q4_K_M | 24.8 GB | 26.7 tok/s | Runs welltight fit |
| 87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking. | Nex N2 Mini (35B-A3B) | Q4_K_M | 24.3 GB | 26.7 tok/s | Runs welltight fit |
| 87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking. | Gemma 4 26B-A4B | Q4_K_M | 18.8 GB | 21.5 tok/s | Runs well |
| 86Excellent quality (A+) · 86 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking. | KAT-Coder V2.5 Dev (35B-A3B) | Q4_K_M | 24.0 GB | 26.7 tok/s | Runs welltight fit |
| 86Excellent quality (A+) · 86 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking. | Sage-A1.5 35B-A3B | Q4_K_M | 24.8 GB | 26.7 tok/s | Runs welltight fit |
| 83Great quality (A) · 83 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking. | GLM-4.7 Flash (30B-A3B) | Q4_K_M | 22.2 GB | 26.7 tok/s | Runs well |
| 85Excellent quality (A+) · 85 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking. | Ornith 1.5 35B-A3B | Q4_K_M | 24.8 GB | 26.7 tok/s | Runs welltight fit |
| 84Great quality (A) · 84 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking. | Apodex 1.1 Mini (35B-A3B) | Q4_K_M | 24.8 GB | 26.7 tok/s | Runs welltight fit |
Best Image models for the QNAP QAI-16700 32.0 GB
| Quality | Model | Quant | Memory | Speed | Runs |
|---|---|---|---|---|---|
| 90Excellent quality (A+) · 90 of 100: how capable the model is at FP8, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. | Krea 2 Turbo | FP8 | 20.2 GB | 3.5 min | Very slow |
| 82Great quality (A) · 82 of 100: how capable the model is at FP8, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. | FLUX.2 [klein] 9B | FP8 | 19.9 GB | 87 s | Slow |
| 78Great quality (A) · 78 of 100: how capable the model is at FP16, whatever it runs on. Match score 81/100 (quality, fit, speed and memory use) sets the ranking. | FLUX.2 [klein] 4B | FP16 | 18.1 GB | 54 s | Slow |
| 87Excellent quality (A+) · 87 of 100: how capable the model is at Q8 GGUF, whatever it runs on. Match score 80/100 (quality, fit, speed and memory use) sets the ranking. | ERNIE-Image | Q8 GGUF | 18.9 GB | 22.6 min | Very slow |
| 85Excellent quality (A+) · 85 of 100: how capable the model is at NF4, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking. | Ideogram 4 | NF4 | 21.8 GB | 24.6 min | Very slow |
| 90Excellent quality (A+) · 90 of 100: how capable the model is at Q8 GGUF, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking. | FLUX 3 Image Klein | Q8 GGUF | 12.8 GB | 5.9 min | Very slow |
Best Video models for the QNAP QAI-16700 32.0 GB
| Quality | Model | Quant | Memory | Speed | Runs |
|---|---|---|---|---|---|
| 83Great quality (A) · 83 of 100: how capable the model is at FP8, whatever it runs on. Match score 73/100 (quality, fit, speed and memory use) sets the ranking. | Bernini-R A14B (video generation + editing) | FP8 | 20.3 GB | 10.1 h | Very slowencoder unloaded after use |
| 74Great quality (A) · 74 of 100: how capable the model is at FP8, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. | HY-OmniWeaving (8.3B, unified video) | FP8 | 24.6 GB | 7.1 h | Very slowtight fit |
| 70Great quality (A) · 70 of 100: how capable the model is at Q4 GGUF, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. | Wan 2.2 S2V 14B (speech to video) | Q4 GGUF | 21.6 GB | 17.9 h | Very slow |
| 68Good quality (B) · 68 of 100: how capable the model is at Q4 GGUF, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. | Wan 2.1 I2V 14B 480P | Q4 GGUF | 20.6 GB | 16.1 h | Very slow |
| 79Great quality (A) · 79 of 100: how capable the model is at Q4 GGUF, whatever it runs on. Match score 70/100 (quality, fit, speed and memory use) sets the ranking. | LTX-2.3 | Q4 GGUF | 24.5 GB | 7.5 h | Very slowtight fit · encoder unloaded after use |
Best Decision models for the QNAP QAI-16700 32.0 GB
| Quality | Model | Quant | Memory | Speed | Runs |
|---|---|---|---|---|---|
| 78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 88/100 (quality, fit, speed and memory use) sets the ranking. | OpenAI Privacy Filter | F16 | 3.8 GB | 41 ms | Runs great |
| 68Good quality (B) · 68 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. | Llama Prompt Guard 2 22M | Q8_0 | 0.7 GB | 95 ms | Runs great |
| 80Great quality (A) · 80 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. | GLiNER2.5 Decide | Q8_0 | 1.2 GB | 318 ms | Runs well |
| 86Excellent quality (A+) · 86 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. | Jina Reranker v3.5 | Q8_0 | 3.6 GB | 725 ms | Slow |
| 84Great quality (A) · 84 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. | Jina Reranker v3 | Q8_0 | 3.6 GB | 725 ms | Slow |
| 66Good quality (B) · 66 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. | GLiClass Base v3 | Q8_0 | 0.8 GB | 126 ms | Runs great |
| 90Excellent quality (A+) · 90 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. | Qwen3 Reranker 4B | Q8_0 | 7.3 GB | 4.8 s | Very slow |
| 74Great quality (A) · 74 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. | GLiClass ModernBERT Large v3 | Q8_0 | 1.1 GB | 260 ms | Runs well |
Other products with the AMD Ryzen 7000 series (desktop, incl. PRO 7645 / 7745) NAS (32 GB)
TS-h3077AFUProduct page (opens in a new tab)
TS-h1077AFUProduct page (opens in a new tab)
TS-h2477AXU-RPProduct page (opens in a new tab)
TS-h1677AXU-RPProduct page (opens in a new tab)
TS-h1277AXU-RPProduct page (opens in a new tab)
HL15 Beast (Fully Built, ASUS ProArt / Ryzen 9)Product page (opens in a new tab)
Spec sources
Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.